We re-examined the issue of active versus passive proprioception to more fully characterize the accuracy afforded by proprioceptive information in natural, unconstrained, movements in 3-dimensions. Subjects made pointing movements with their non-dominant arm to various locations with eyes closed. They then proprioceptively localized the tip of its index finger with a prompt pointing movement of their dominant arm, thereby bringing the two indices in apposition. Subjects performed this task with remarkable accuracy. More remarkably, the same subjects were equally accurate at localizing the index finger when the arm was passively moved and maintained in its final position by an experimenter. Two subjects were also tested with eyes open, and they were no more accurate than with eyes closed. We also found that the magnitude of the error did not depend on movement duration, which is contrary to a key observation in support of the existence of an internal forward model-based state-reconstruction scheme. Three principal conclusions derive from this study. First, in unconstrained movements, proprioceptive information provides highly accurate estimates of limb position. Second, so-called active proprioception does not provide better estimates of limb position than passive proprioception. Lastly, in the active movement condition, an internal model-based estimation of limb position should, according to that hypothesis, have occurred throughout the movement. If so, it did not lead to a better estimate of final limb position, or lower variance of the estimate, casting doubt on the necessity to invoke this hypothetical construct.